Add ContextPropagation runtime util + captureContext operator

[simonbasle]

In this PR we introduce a ContextPropagation runtime-detection utility class
as well as a contextCapture() operator. If context-propagation isn't on the
classpath, this operator is NO-OP.

If context-propagation is on the classpath however, the operator will use
it to capture the current context during the subscription phase (at the point
a contextWrite would be effected) and store it in the ContextView visible
from upstream of the operator.

TODO:

• Mono version
• mention in reference guide? whole section?
• marble diagram

[simonbasle]

as there is currently no standard marble diagram representation of the Context, we won't provide a marble diagram for now

[rstoyanchev]

Looks good. Some minor comments on the docs.

[simonbasle]

tentative commit message:

This commit introduces a ContextPropagation runtime-detection utility
class as well as a `contextCapture()` operator.

If context-propagation isn't on the classpath, this operator is NO-OP.

If context-propagation is on the classpath however, the operator will
use it to capture the current context during the subscription phase
(at the point a contextWrite would be effected) and store it in the
ContextView visible from upstream of the operator.

[chemicL]

Same remarks as in case of Flux.

[chemicL]

cc @marcingrzejszczak @rstoyanchev

[chemicL]

Approved, although I left a few comments - some are minor, for the docs; one I believe to be important for UX, but possible to improve as a follow-up if the logging option is chosen. With additional method, it's probably best to introduce now.

[simonbasle]

looks like there was no strong consensus on the contextCapture operator being NO-OP if the lib is not on the classpath.
so far all the features in reactor-core that depend on an optional dependency have been exposed that way (eg. Flux#metrics()).

I've fixed the comments on the refguide part.

Note that this PR is conflicting with #3215, as it adds usage of context-propagation including methods that have seen breaking changes in latest snapshots... this is just out-of-sync with M5

[simonbasle]

I'll need to rebase against current main (which depends on context-propagation M5)